Describe the bug
Earlier, support for LOG and ROUND scalar functions with two arguments was added in https://github.com/cube-js/arrow-datafusion/pull/135. However, the second argument type was specified as Int32. It allows it to match columns but prevents from matching integer literals.
To Reproduce
Query generated by QuickSight:
SELECT
date_trunc(\'day\', "period") AS period_tg",
SUM("clicks") AS clicks_sum",
ROUND(CAST(CAST(SUM("clicks") - (MIN(SUM("clicks")) OVER ()) AS FLOAT) / NULLIF(CAST((MAX(SUM("clicks")) OVER ()) - (MIN(SUM("clicks")) OVER ()) AS FLOAT), 0) AS NUMERIC), 3) AS "68c185d7-907c-3bb3-97f2-ec592e580dc0"
FROM "public"."revenue"
GROUP BY date_trunc(\'day\', "period")
ORDER BY date_trunc(\'day\', "period") NULLS FIRST
LIMIT 500
Casting the second argument by doing round(3.142, cast(1 as integer)) works but since the BI tool is generating the query, there no actual control over the second argument.
Expected behaviorLOG and ROUND should work with integer literals as the second argument. They are indeed matched as Int64 values, as confirmed by @MazterQyou.
Describe the bug Earlier, support for
LOGandROUNDscalar functions with two arguments was added in https://github.com/cube-js/arrow-datafusion/pull/135. However, the second argument type was specified asInt32. It allows it to match columns but prevents from matching integer literals.To Reproduce Query generated by QuickSight:
Casting the second argument by doing
round(3.142, cast(1 as integer))works but since the BI tool is generating the query, there no actual control over the second argument.Expected behavior
LOGandROUNDshould work with integer literals as the second argument. They are indeed matched asInt64values, as confirmed by @MazterQyou.Version: 1.1.0
Additional context Slack thread: https://cube-js.slack.com/archives/C01FU5AP9LJ/p1727956303761059